Many large biomolecules contain highly flexible structural components that undergo large-scale anisotropic and collective deformations. Ideally, these deformations should be more accurately described using anisotropic temperature B-factors. However, very frequently, large complexes containing highly flexible components yield crystals that only diffract to limited resolutions (3~4.5[unreadable]). Thus, limited by the relatively small number of unique reflections, a full-scale conventional anisotropic refinement that requires three positional and six thermal parameters for each atom is impractical for many such systems. As a result, they are often refined with one isotropic B-factor for each atom at the best scenario. The inability to model these anisotropic deformations with reasonable accuracy in turn deteriorates the refinement of positional parameters, slows down the overall convergence, and results in large errors in refined structural models. Therefore, new methods are urgently needed to cope with large deformations of protein structures in structure determination and functional study. Hypotheses: Large-scale deformations of biomolecules contribute significantly to the errors in structure determination, which can be reduced by anisotropic refinement using a small number of collective normal modes. General Objectives: Our focus has been on developing new simulation methods to represent more realistically and efficiently large-scale deformations of biomolecules in structure determination and functional study. In this funding cycle, a new normal-mode-based X-ray refinement protocol (NM-XREF) will be developed and tested in a large set of limited-resolution structures. Specific Aims: 1) Algorithmic and software development. New algorithmic development will be pursued to improve the efficiency and accuracy of NM-XREF. Furthermore, substantial efforts will be invested to develop the NM-XREF protocol into a user-friendly software package for serving the entire structural biology community. 2) Systematic benchmark of NM-XREF. We will systematically test NM-XREF and compare it with TLS on over 50 biomolecular systems. The outcome is expected to provide a general guideline for the application of NM-XREF. 3) Refinement of a selected group of biological systems. We have selected some of the most challenging biological systems for a more thorough investigation through multiple cycles of NM-XREF refinement and manual adjustment. The final structural models are expected to provide new insights into the functionally important dynamics of the systems. 4) Structure determination of mammalian fatty acid synthase. By using NM-XREF, we hope to resolve some of the mobile structural components missing in previous studies. Our extensive preliminary results suggest that, for a large number of limited-resolution structures refined using conventional methods, the improvement of model quality by NM-XREF is still substantial. Moreover, NMXREF not only outperforms the TLS method, but also maximizes the gain by TLS when they are sequentially utilized in some cases. Thus, it is of a high priority to develop NM-XREF into a friendly tool for the community. PUBLIC HEALTH RELEVANCE Atomic structures of biomolecules are critical to the understanding of their cellular functions, which often involve large-scale conformational deformations, especially for large protein assemblies. Although functionally important, those large-scale deformations impose enormous difficulties on structural refinement in X-ray crystallography. This proposal aims to develop a new X-ray refinement protocol that, with fewer refinement parameters, provides a more accurate description of conformational deformations in structure determination at limited resolutions (3~4.5[unreadable]).